-
Notifications
You must be signed in to change notification settings - Fork 1.1k
Minutes Data Working Group 16 Aug 2022
Brad edited this page Aug 16, 2022
·
2 revisions
- (Brad) Review of Data WG tagged items in the MONAI forums
- Data WG Issues
- Data WG Discussions
- Any MONAI Advisory Board topics
- (Carole) Lead discussion on “Access and storage of data” from Evaluation and Benchmarking Workgroup
- (Brad) Present draft survey and get approval for its use from the committee
- (Group) Review Data WG parking lot items
- Exploration of data-related issues arising from MONAI Label and Deploy
- E.g., one question that has come up : annotation representation of videos
- Exploration of data-related issues arising from MONAI Label and Deploy
No new issues or discussions posted.
3. (Brad) Present draft survey and get approval for its use from the committee
MONAI Steering Committee may be building a general survey where we could tack questions on; Brad will join next steering call to get an update.
- Recurring conversations in the community regarding what data formats, modalities, and use cases MONAI can be used for
- We need improved accessibility on data that's supported
- Something on the website, with details for
- What's supported
- What's not officially supported but there are examples are out there
- What's not yet supported (and maybe good for community input)
- E.g., Skin Cancer Training using MONAI
- May include creative steps on making it work (e.g. while streams may not be supported, video could be divided into a series of frames)
Action Item:
- (Brad-DONE!) Start document to gather formats, modalities, use cases
- (All) Add examples from the community, web, etc.
- (Brad) Talk to Prerna, Michael to discuss where such a document lives (in docs, on main pages, etc)
2. (Carole) Lead discussion on “Access and storage of data” from Evaluation and Benchmarking Workgroup
- How do we help support making challenge data reproducible and consistent for MONAI users?
- e.g., making fetching the datasets easier, in the right structures to make working with MONAI easy?
- So that challenge participants won't have to care about how to access the data
- What help can we provide help to challenge organizers?
- We would need a document of guidelines, best practices, or suggested structures that we could give to challenge organizers
- Challenge providers could include links to "MONAI-compatible structure"
- Do we list challenges that have MONAI-compatible structures on a MONAI website? E.g., so that visitors to MONAI can learn about challenges that they can get started quickly. This would be mutually beneficial (traffic to the challenge from MONAI + traffic to MONAI from the challenges)
- What could we learn from winners of challenges who used MONAI, how much work did they do to get the data in the right format?
- Are there ways that we bring the experience of winners back to MONAI for learnings for the future?
- Maybe we should create a survey that they can fill out over time?
- Questions around preparation of the data itself, and carried all the way through to MONAI Zoo?
- Asking, e.g., what did you use that made it happen in the end?
- IDC sample Jupyter notebook: exploring_segmentations_for_monai
- Four chapters, covering 1. Access, 2. Download, 3. Constraining and 4. Conversion
- Dataset is hosted on Google
- A good "chapter 5" would be extending this to include MONAI training steps
- Such a guide already exists for TCIA: Quick Start With TciaDataset
- Could this be used to front load into a database like XNAT to then be used to feed MONAI?
- Makes sense - but there's still a bridge to get the data into a format for MONAI
- What modules in MONAI is needed to streamline this? E.g., quality control modules?
- Have to strike the balance between what MONAI code does and what users of MONAI do
- Structures is flexible, we need to provide guidance to make this work